1. The Election Integrity Partnership
to tag platform partners on a ticket for action. They also communicated
with the EIP’s partners in government, and could request further infor-
mation from election officials if necessary. Once a ticket reached Tier 3,
the manager decided whether to put it into a holding queue for ongoing
monitoring, assign the ticket back to a Tier 2 analyst to produce a public
blog post or Twitter thread discussing the issue, or close a ticket if it had
been resolved.
Team members from each of these tiers were divided into on-call shifts. Each
shift was four hours long and led by one on-call manager. It was staffed by a
mix of Tier 1 and Tier 2 analysts in a 3:1 ratio, ranging from five to 20 people.
Analysts were expected to complete between two to five shifts per week. The
scheduled shifts ran from 8:00 am to 8:00 pm PT for most of the nine weeks
of the partnership, ramping up only in the last week before the election from
12-hour to 16- to 20-hour days with all 120 analysts on deck.
A note on fact-checking: the EIP was not a fact-checking organization, and in
preliminary assessments of whether an event in a ticket was potentially misin-
formation, analysts first looked to the work of others. One of the complexities
related to misleading information is that it is not always possible to verify the
claims; professional fact-checkers confronted with these situations may use
labels like “inconclusive” or “partially true” to convey uncertainty where it exists.
Where possible, our analysts identified an external fact-checking source from
news sites, credible fact-checking organizations, or statements from a local
election official when filing tickets. Analysts also used open source investigation
techniques, such as reverse image searches or location identifications, to de-
termine if images or videos tied to an incident were taken out of their original
context. Our analysts identified at least one external fact-check source for
approximately 42% of the in-scope tickets. For some tickets, it was not possible
to find an external fact-check for the content, either because no fact-checker
had yet addressed the issue, or because the information was resistant to simple
verification—for example, content based on unconfirmed or conflicting claims
from a whistleblower, conspiracy theories that claimed invisible forces at work,
and narratives based on factual claims (e.g., discarded ballots) but spread within
misleading frames that exaggerated the
